01Five facts, written out on one sheet
Before you open any form, write these five things down in plain sentences and keep the sheet beside you for the whole submission. Every document in the packet that touches one of them has to say it the same way, in numbers that match and in words a participant would recognise as describing the same study.
- How many people, given as one figure or one range and never as two different ranges
- Who is eligible, and just as importantly who is excluded and on what basis
- What a participant actually does, step by step, and how long it takes them
- Where the data sits, in what form, for how long, and when it is destroyed
- Who may see it, in identifiable or de-identified state, and under what conditions
02How the drift actually happens
A protocol says twenty to thirty participants. The recruitment notice says about twenty-five. The consent form says approximately thirty. Nothing dishonest has occurred. Three documents were written on three different days by one person who knew perfectly well what they meant, and the board is now holding a set that describes three studies.
The second drift is worse because it is invisible to the author. An instrument gains two questions late in the drafting, and the duration promised in the consent form stops being true. The consent form is the document a participant relies on when they agree, so a board treats an inaccurate one seriously. Change the instrument, and you have changed the consent form, the notice and the protocol whether you edited them or not.
A third one hides in the contact block, which nobody proofreads. The protocol names your chair as responsible investigator, the consent form offers your personal mobile number, and the recruitment notice gives a departmental address at the site. Somebody who wants to withdraw now has three routes and no way to tell which one is real. Settle it once: who is contacted for a question about the study, who for withdrawal, and who for a concern about how it is being run.
03What the board is doing, and what it is not
It is not peer review. Nobody on it is deciding whether your question is interesting, whether the design is the strongest available, or whether the topic will make a good chapter. The remit is narrower and it is worth taking seriously on its own terms: risk to the people taking part, and whether the consent they give is genuinely informed.
Two consequences follow. Do not classify your own submission. Describe the procedures accurately and let the board decide which review category applies, because a submission that declares itself exempt and is not reads as a request rather than a description. And check your certification currency before anything else; a lapsed human-subjects certificate, CITI or otherwise, halts a packet before a single scientific sentence gets read.
04The pass to run the night before filing
Read the consent form first, as a participant would, slowly, out loud. Then read the protocol with the five-fact sheet beside you and tick each fact where it appears. Then the recruitment material, which people forget is part of the packet and which the board reads for pressure and for promises. Then work through the instrument yourself with a clock running, because the duration you have promised is now a testable claim.
Finish on the boring layer: dates, version numbers, file names, and whether the signature blocks belong to the people who will actually sign them. A packet returned for a version mismatch costs the same waiting time as one returned for a substantive concern, and it is the kind of return that is entirely yours to prevent.
Where a protocol has already been through one revision cycle and is still coming back, the fault is usually structural rather than clerical, and that is the work IRB clearance takes on.
FAQQuestions to control
How long should I expect to wait for a first decision?
Longer than the calendar in your head, and the figure varies by board and by review category, so take it from your own institution rather than from anyone else. Plan your data collection from the date of an approval letter you do not yet have, not from the date you filed. Every schedule that assumes approval on the first pass eventually costs a term.
Does a small student project still need review?
That determination belongs to the board, not to you or your chair. Submit a description and let them classify it, because the cost of asking is a form and the cost of guessing wrong is data you are not permitted to use. Some projects genuinely fall outside the remit, and a written answer saying so is worth having in your appendix.
What does a revision letter actually mean?
Usually that a reviewer could not confirm something from the documents in front of them. Read each item as a question rather than a criticism, answer it in the place the reviewer was looking rather than in a cover note, and change every other document the answer touches. Replying to one item while leaving a contradiction elsewhere is what produces a second letter.
Can somebody else write my protocol for me?
The design and the answers have to be yours, because you are the person who will be held to them and asked about them at a defence. What can be handled alongside you is the drafting discipline: turning your design into the board's language, running the consistency audit across the whole packet, and rebuilding a revision response so that every touched document changes together.
What if I have already gathered some of the data?
Stop, and put it in the submission rather than around it. Approval does not reach backwards, so material collected before an approval letter usually cannot be used however carefully it was collected. Tell your chair the same day and let the board decide what happens to it. The version of this you can recover from is one you declared; the version you cannot is one somebody else finds.
